Spatial Educational Intelligence: Grounding Learning in Three-Dimensional Environments
Abstract
Artificial intelligence in education has advanced from rule-based tutoring systems and digital learning platforms to conversational agents, generative artificial intelligence, and multimodal foundation models. However, educational intelligence is frequently organized around the interpretation of textual, visual, or symbolic information, while the spatial structure of the learning environment remains secondary to the system's reasoning and decision-making processes. In actual educational settings, learning depends not only on what students know, but also on where students are located, how they interact with one another, which materials and instruments are accessible, how objects are spatially arranged, and how these relationships change during instructional activities. Meanwhile, spatial intelligence has emerged as an important research direction in artificial intelligence, encompassing three-dimensional perception, geometric reasoning, spatial representation, scene understanding, and spatially informed action. This paper introduces and formally defines Spatial Educational Intelligence (SEI) as a proposed research framework for artificial intelligence systems that construct and update spatial representations of educational environments, connect spatial states with educational objectives, predict the consequences of spatially situated instructional decisions, and select or recommend educational actions according to spatial, pedagogical, and safety constraints. The paper distinguishes SEI from Artificial Intelligence in Education (AIED), spatial intelligence, educational visualization, augmented and virtual reality learning, educational robotics, and Embodied Educational AI (EEAI). A five-layer SEI architecture is proposed, comprising Spatial Perception, Educational-Spatial Representation, Spatial-Educational Reasoning, Educational Action Selection, and Outcome Verification and Adaptation. A central mechanism, termed Educational-Spatial Grounding (ESG), is introduced to connect educational intentions with the spatial conditions, relationships, constraints, and consequences relevant to their implementation. The paper further proposes a taxonomy of 60 potential SEI applications across classroom orchestration, collaborative learning, mathematics, science, laboratory education, vocational training, programming, accessibility, physical education, assessment, educational administration, and spatially situated tutoring. A spatially adaptive science classroom is presented as a conceptual prototype illustrating how SEI could support instructional decisions without requiring every system to possess a physical body. The paper argues that the central research transition is from AI that represents educational knowledge independently of its spatial context to AI that uses spatial context as an explicit basis for educational reasoning and decision-making. SEI therefore provides a conceptual framework for connecting artificial intelligence, spatial computing, learning sciences, multimodal perception, educational robotics, and adaptive learning environments. The framework is conceptual and requires empirical validation before its effectiveness can be established.
Related
- Intelligent educational agent-based collaborative scaffolding for STEM interdisciplinary project-based learning: a multi-dimensional framework and empirical evaluation
- Educational Digital Twins: A Closed-Loop Framework for Modeling and Transforming Learning Environments
- Sovereign-OCR: Zero-Hallucination Neuro-Symbolic Document Intelligence with Self-Healing CMap Recovery, Active Ontology Grounding, and Deterministic RDF 1.2 Provenance
- Generative artificial intelligence use and interpersonal educational interaction among nursing students: self-regulated learning as a statistical mediator
- Panpsychism: A Philosophical Overview Without Mathematical or Empirical Grounding — E8 Intelligence Research
BibTeX
@misc{zhang2026spatial,
title = {{Spatial Educational Intelligence: Grounding Learning in Three-Dimensional Environments}},
author = {Jincheng Zhang},
howpublished = {Zenodo},
year = {2026},
month = oct,
doi = {10.5281/zenodo.23592696},
url = {https://zenodo.org/records/23592696},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Spatial Educational Intelligence: Grounding Learning in Three-Dimensional Environments
AU - Jincheng Zhang
PY - 2026
DA - 2026/10/23
PB - Zenodo
DO - 10.5281/zenodo.23592696
UR - https://zenodo.org/records/23592696
AB - Artificial intelligence in education has advanced from rule-based tutoring systems and digital learning platforms to conversational agents, generative artificial intelligence, and multimodal foundation models. However, educational intelligence is frequently organized around the interpretation of textual, visual, or symbolic information, while the spatial structure of the learning environment remains secondary to the system's reasoning and decision-making processes. In actual educational settings, learning depends not only on what students know, but also on where students are located, how they interact with one another, which materials and instruments are accessible, how objects are spatially arranged, and how these relationships change during instructional activities. Meanwhile, spatial intelligence has emerged as an important research direction in artificial intelligence, encompassing three-dimensional perception, geometric reasoning, spatial representation, scene understanding, and spatially informed action. This paper introduces and formally defines Spatial Educational Intelligence (SEI) as a proposed research framework for artificial intelligence systems that construct and update spatial representations of educational environments, connect spatial states with educational objectives, predict the consequences of spatially situated instructional decisions, and select or recommend educational actions according to spatial, pedagogical, and safety constraints. The paper distinguishes SEI from Artificial Intelligence in Education (AIED), spatial intelligence, educational visualization, augmented and virtual reality learning, educational robotics, and Embodied Educational AI (EEAI). A five-layer SEI architecture is proposed, comprising Spatial Perception, Educational-Spatial Representation, Spatial-Educational Reasoning, Educational Action Selection, and Outcome Verification and Adaptation. A central mechanism, termed Educational-Spatial Grounding (ESG), is introduced to connect educational intentions with the spatial conditions, relationships, constraints, and consequences relevant to their implementation. The paper further proposes a taxonomy of 60 potential SEI applications across classroom orchestration, collaborative learning, mathematics, science, laboratory education, vocational training, programming, accessibility, physical education, assessment, educational administration, and spatially situated tutoring. A spatially adaptive science classroom is presented as a conceptual prototype illustrating how SEI could support instructional decisions without requiring every system to possess a physical body. The paper argues that the central research transition is from AI that represents educational knowledge independently of its spatial context to AI that uses spatial context as an explicit basis for educational reasoning and decision-making. SEI therefore provides a conceptual framework for connecting artificial intelligence, spatial computing, learning sciences, multimodal perception, educational robotics, and adaptive learning environments. The framework is conceptual and requires empirical validation before its effectiveness can be established.
ER -